The ChinaHighPM2.5 data set: generation, validation, and spatiotemporal variations from 2000 to 2018 in China

Abstract:

Fine particulate matter (PM_2.5) can significantly affect the atmospheric environment and harm human health. Current satellite-derived PM_2.5 data have overall low accuracies with coarse spatial resolutions limited by data sources and models. Ground-based PM_2.5 records only date back to 2013 in China, while air pollution level has experienced the most dramatic changes over a couple of decades. To reveal the spatiotemporal variations of PM_2.5, long-term, high-quality, and high-spatial-resolution aerosol optical depths retrieved by the MODIS MAIAC algorithm were employed to estimate PM_2.5 concentrations using a newly developed ensemble machine learning method, i.e., Space-Time Extra-Trees (STET) model. Our model can capture well variations in PM_2.5 concentrations at spatiotemporal scales at higher accuracies (i.e., CV-R² = 0.85–0.90) and stronger predictive powers (i.e., R² = 0.78–0.81) than previously reported. The resulting PM_2.5 dataset for China (ChinaHighPM_2.5) provides the longest records (2000 to 2018) with the highest quality at a 1 km spatial resolution. In general, PM_2.5 concentrations showed increasing trends around 2007 and remained high until 2013, after which declined substantially, thanks to a series of government actions combating air pollution in China. While the nationwide PM_2.5 concentrations have decreased by 0.89 μg/m³/yr (p < 0.001) during the last two decades, of the reduction has accelerated to 4.08 μg/m³/yr (p < 0.001) in the last six years, indicating a significant improvement in air quality. The ChinaHighPM_2.5 data set will enable more insightful analyses regarding the causes and attribution of pollution over the medium- or small-scale areas.